Brain MRI Classification & Clustering
CompletedMay 2026
End-to-end ML pipeline on 6,056 brain MRI images with a 55-feature vector; XGBoost reached 91.67% accuracy (AUC 0.985).
About
An end-to-end ML pipeline on 6,056 brain MRI images (Glioma, Meningioma, Pituitary) for the COMP 466 — Business Intelligence course, supervised by Prof. Dr. Adem Karahoca.
- Engineered a 55-feature vector using GLCM, Hu moments, HOG and FFT descriptors
- Benchmarked 7 classifiers and 4 clustering algorithms
Results
- Best classifier: XGBoost — 91.67% accuracy, AUC 0.985
- Best clustering: Agglomerative (average linkage) — Silhouette score 0.626
Tech
- Python
- scikit-learn
- XGBoost
- GLCM
- HOG
- FFT